Gelsolin binding to phosphatidylinositol 4,5-bisphosphate is modulated by calcium and pH.

The actin cytoskeleton of nonmuscle cells undergoes extensive remodeling during agonist stimulation. Lamellipodial extension is initiated by uncapping of actin nuclei at the cortical cytoplasm to allow filament elongation. Many actin filament capping proteins are regulated by phosphatidylinositol 4,5-bisphosphate (PIP2), which is hydrolyzed by phospholipase C. It is hypothesized that PIP2 dissociates capping proteins from filament ends to promote actin assembly. However, since actin polymerization often occurs at a time when PIP2 concentration is decreased rather than increased, capping protein interactions with PIP2 may not be regulated solely by the bulk PIP2 concentration. We present evidence that PIP2 binding to the gelsolin family of capping proteins is enhanced by Ca2+. Binding was examined by equilibrium and nonequilibrium gel filtration and by monitoring intrinsic tryptophan fluorescence. Gelsolin and CapG affinity for PIP2 were increased 8- and 4-fold, respectively, by microM Ca2+, and the Ca2+ requirement was reduced by lowering the pH from 7.5 to 7.0. Studies with the NH2- and COOH-terminal halves of gelsolin showed that PIP2 binding occurred primarily at the NH2-terminal half, and Ca2+ exposed its PIP2 binding sites through a change in the COOH-terminal half. Mild acidification promotes PIP2 binding by directly affecting the NH2-terminal sites. Our findings can explain increased PIP2-induced uncapping even as the PIP2 concentration drops during cell activation. The change in gelsolin family PIP2 binding affinity during cell activation can impact divergent PIP2-dependent processes by altering PIP2 availability. Cross-talk between these proteins provides a multilayered mechanism for positive and negative modulation of signal transduction from the plasma membrane to the cytoskeleton.

Phosphoinositides are important in signal transduction, both as precursors to signaling molecules and as physical anchors and regulators of proteins (1,2). Among these, the D4 phosphoinositide, phosphatidylinositol 4,5-bisphosphate (PIP 2 ), 1 has been implicated as a potential mediator of actin cytoskeletal rearrangements (3,4). PIP 2 modulates many actin regula-tory proteins. These include the following: actin severing and/or capping proteins (gelsolin (5), CapG (6), and capping protein (also known as Cap Z) (7)), monomer-binding proteins (profilin (8) and cofilin (9)), and other actin-binding proteins (␣-actinin (10) and vinculin (11)). It has been hypothesized that PIP 2 induces explosive actin assembly by dissociating capping proteins from filament ends and releasing actin monomers from actin-sequestering proteins (3,7,12). The involvement of PIP 2 in actin polymerization is supported by recent experiments that show that Rac1 and RhoA, monomeric GTPases of the Rho family that have well defined effects on the cytoskeleton (13), stimulate the synthesis of PIP 2 (14 -16). Furthermore, manipulations that alter the availability of PIP 2 in cells have profound effects on agonist and/or Rac1-induced filament end capping, actin polymerization, and cell motility (16,17). However, although the time courses of PIP 2 hydrolysis and recovery correlate in some cells (16,18), they do not in most of the cells examined (19 -21). Particularly puzzling is the finding that, in many cells, actin polymerizes at a time when PIP 2 level is reduced, rather than increased, as would be expected if uncapping and monomer desequestration are initiated by PIP 2 . To explain this discrepancy, it is often hypothesized that local PIP 2 availability can be enhanced by compartmentalization or differential turnover (22)(23)(24), even as the bulk PIP 2 mass is reduced. The equally attractive possibility that PIP 2 binding is regulated by signals generated during agonist stimulation has not been considered.
Agonist-stimulated cells exhibit complex Ca 2ϩ oscillations and pH transients. These signals alter the binding of gelsolin and CapG to actin, by inducing a conformational change (6,(25)(26)(27). In this study, we tested the effect of Ca 2ϩ and pH on the binding of the gelsolin family proteins to PIP 2 and found that they affect PIP 2 binding in an interdependent manner. We identified the domains in gelsolin that impart such regulation and elucidated the relation between the NH 2 -terminal and COOH-terminal halves of the protein. Since gelsolin modulates the activity of many PIP 2 -regulated proteins with important signaling functions in vivo (28) and in vitro (29 -31), our results have important implications for how the gelsolin family proteins are regulated during agonist signaling and how the activity of other PIP 2 -dependent cytoskeletal and noncytoskeletal proteins can be coordinated.

EXPERIMENTAL PROCEDURES
Expression and Purification of Recombinant CapG, Gelsolin, and Gelsolin Domains-Gelsolin has six semihomologous domains (S1-6), which can be further divided into two functional halves (32). The expression vectors for the gelsolin NH 2 -terminal half (S1-3), gelsolin S1, gelsolin S2-3, and CapG have been described previously (33)(34)(35). The full-length gelsolin expression vector (encompassing the entire human plasma gelsolin coding sequence) was constructed by ligating gelsolin cDNA to pet3a via the BamHI site. Recombinant proteins were expressed in bacteria and purified using sequential anion and cation exchange chromatography (34). Protein concentration was determined by the method of Bradford (36), and protein purity was assessed by SDS-polyacrylamide gel electrophoresis.
The COOH-terminal half expression vector was constructed by using polymerase chain reaction to generate a fragment encompassing human plasma gelsolin nucleotides 1298 -1753. The forward primer contains a XhoI site (ACC TCC ACT CTC GAG GCC GCC), and the reverse primer has a SmaI site (CAA CAG CCC GGG TGG CT). The polymerase chain reaction product was cloned into Bluescript KSϩ via the XhoI/ SmaI sites. This construct was digested with SmaI and blunt endligated with a downstream gelsolin fragment. The fragment was excised with BamHI from full-length gelsolin cDNA in Bluescript KSϩ (gelsolin SmaI site at nucleotide 1750 and vector multiple cloning SmaI site downstream of the termination codon). The resultant cDNA was digested with SpeI (in the 3Ј multiple cloning region, downstream of SmaI) and filled in with CT nucleotides to create a site with a two-base overhang compatible with that of HindIII. The other end was released by digestion with XhoI and ligated to PGEX K6 vector that was linearized with HindIII (site partially filled in with nucleotides AG to generate a two-base overhang compatible with the partially filled in SpeI) and XhoI. The fusion protein contained a 30-kDa GST followed by a 40-kDa gelsolin COOH-terminal half. The COOH-terminal gelsolin was cleaved from GST bound to a column with thrombin.
Phospholipid-PIP 2 was purchased from Calbiochem. Micelles were prepared by dissolving the dried lipid in water to a final concentration of 2 mg/ml and sonicating for 5 min. at maximum power (model W185; Heat Systems Ultrasonics, Inc., Farmingdale, NY). Large unilamellar vesicles at a 5:1 phosphatidylcholine:PIP 2 ratio were made with an extruder (Lipex Biomembranes, Vancouver, Canada) as described by Machesky et al. (37).
Small Zone Gel Filtration-The assay was similar to that described previously for studying lipid binding to most actin regulatory proteins (33,35,38). This is because small proteins bound to PIP 2 micelles or mixed vesicles migrate faster than the unbound proteins. Proteins were incubated with lipid for 30 min at room temperature, and 100 l of the mixture was chromatographed at 4°C through a Superdex 75 HR 10/30 column (Pharmacia Biotech Inc.), equilibrated with pH 7.0 or 7.5 buffers containing 25 mM Hepes, 100 mM KCl, 0.5 mM ␤-mercaptoethanol, 0.4 mM EGTA with or without CaCl 2 . Lipid was not included in the elution buffer. Fractions were eluted at 0.5 ml/min, and 0.5-ml fractions were collected. The elution profile was monitored by absorbance at 280 nm. The amount of unbound protein was determined from the protein absorbance peak. The lipid-bound protein was calculated as the difference between the total protein applied minus the unbound protein. The apparent dissociation constant (K d ) was calculated as follows.
where r is the ratio of protein bound to each PIP 2 molecule at a given PIP 2 concentration and B max is the maximum number of protein bound per PIP 2 at saturation. Quenching of Intrinsic Tryptophan Fluorescence-Fluorescence spectra were recorded at 30°C with a QM-1 fluorometer (Photon Technology International, Canada). 2 ml of a protein solution (0.3 M, 30°C) in 25 mM Hepes, 100 mM KCl, 0.4 mM EGTA, 0.5 mM ␤-mercaptoethanol, pH 7.5, with or without 36 M free Ca 2ϩ were placed in a 1-cm square quartz cuvette and stirred with a minimagnetic stirrer. After allowing 5 min for equilibration, the tryptophan fluorescence spectrum was recorded by excitation at 292 nm. The excitation and emission beam slits were set at 3 and 2 nm bandwidth, respectively. PIP 2 micelles (at final PIP 2 concentrations ranging from 0.042 to 32.3 M, depending on the protein studied) were added at 2-l increments, and the fluorescence spectra were recorded 5 min after each addition. The total volume of micelles added did not exceed 2% of the initial protein solution volume. The decrease in fluorescence emission at 320 nm was plotted as a function of PIP 2 concentration, and the fluorescence change was assumed to be proportional to the concentration of the protein-phosphoinositide complex. Data were analyzed as described by Ward (40). The apparent dissociation constant, K d , was calculated using the equation, where ⌬F is the fluorescence quenching at a given PIP 2 concentration, ⌬F max is the total fluorescence quenching of the protein saturated with ligand, and [lipid T ] is the concentration of PIP 2 . ⌬F max is estimated by curve fitting of the binding data using the Hyperbol.fit program in SigmaPlot. Alternatively, the intrinsic association constant (K a ) as well as the stoichiometry of binding (p) can be derived using the graphical method of Stinson and Holbrook (41), where is the fractional binding (⌬F/⌬F max ), p is the stoichiometry of binding, [lipid T ] is the total concentration of PIP 2 , and [protein T ] is the total acceptor concentration. When 1/(1 Ϫ ) is plotted against lipid T /, a straight line with a slope of K a and an intercept of protein T / is obtained. The stoichiometry of interaction (p) can be calculated by dividing the intercept with the protein concentration.

Measurement of Free Ca 2ϩ Concentration and pH-
The concentrations of free Ca 2ϩ in EGTA containing solutions with varying amounts of Ca 2ϩ were measured with Ca 2ϩ -sensitive dyes. 5 M Fura-2 was used to determine Ca 2ϩ concentrations below 1 M. Free Ca 2ϩ concentration was calculated (26) assuming the K d of the Fura-2-Ca 2ϩ complex is 229 nM at pH 7.0 and 144 nM at pH 7.5. Calcium green 5N (Molecular Probes, Eugene, OR) was used to measure Ca 2ϩ concentrations higher than 1 M, and free Ca 2ϩ concentration was calculated assuming a K d of 14 M.

RESULTS
CapG Binding to PIP 2 -Small zone gel filtration analyses showed that CapG bound to PIP 2 micelles in a dose-dependent manner. Micelle-bound CapG eluted in the void volume that was well separated from the free protein peak (Fig. 1). Binding to phosphatidylcholine-PIP 2 vesicles gave similar results (data not shown), suggesting that micelles could be used to assess binding, although it is not a physiological substrate. To facilitate comparison under different binding conditions and between different proteins, we attempted to calculate a K d . Equilibrium binding studies suggest that each CapG binds two PIP 2 molecules (see below). Assuming this stoichiometry, the apparent K d for binding to PIP 2 micelles (calculated using Equation  (Table I). These values represent the upper limit, since measurements were not made under equilibrium conditions.
To determine if there is indeed a Ca 2ϩ -induced change, equilibrium binding studies based on the quenching of CapG intrinsic tryptophan fluorescence by PIP 2 were performed. This method has been used to study the binding of profilin (43), phospholipase C␦ (44), and dynamin pleckstrin homology domain (45) to PIP 2 . CapG had an emission maximum of 327 nm, and 36 M Ca 2ϩ produced a small reduction in fluorescence intensity (the ratio of peak fluorescence in EGTA/Ca 2ϩ is 0.92 Ϯ 0.05 (mean Ϯ S.E., n ϭ 5) (Fig. 2, A and B). PIP 2 induced a dose-dependent and saturable decrease in intrinsic fluorescence, without shifting the emission maximum. Micelles alone without CapG did not have significant emission (data not shown). A plot of CapG fluorescence quenching versus PIP 2 concentration showed that saturation was reached at a lower PIP 2 concentration in the presence of Ca 2ϩ than in EGTA (Fig.  3A). The K d values for binding at pH 7.5, calculated according to Equation 3, were 31.9 and 8.4 M in EGTA and Ca 2ϩ , respectively, for the experiment shown. Similar values (24.4 and 6.0 M) were obtained when the data were analyzed using Equation 4 (Table I). These K d values were 3-4 times lower than the small zone gel filtration values, suggesting that CapG-PIP 2 complexes dissociate during nonequilibrium gel filtration. Using a similar protocol, human platelet profilin binds PIP 2 with a K d of 35 M (43), and binding is not affected by Ca 2ϩ .
The stoichiometry of CapG binding was 1.7 in either Ca 2ϩ or EGTA (Table I). Since CapG has one known PIP 2 binding site (6,33), this site appears to bind two PIP 2 molecules. The two PIP 2 bound independently and noncooperatively, as indicated by the Hill coefficients of close to 1 (1.02 Ϯ 0.05 and 1.09 Ϯ 0.02 in EGTA and Ca 2ϩ , respectively) (Fig. 3B, Table I). The exact meaning of this stoichiometry is not clear, because each micelle contains multiple PIP 2 and CapG can potentially bind more than one micelle. Nevertheless, the calculated stoichiometry is  useful for comparison among different proteins. Equilibrium gel filtration validated the K d derived by fluorescence titration. The column was preequilibrated with CapG, and PIP 2 incubated with CapG in the equilibrating buffer was added. The column was then developed with CapG containing equilibration buffer. CapG bound to PIP 2 migrated faster, increasing the CapG content above the equilibration level (peak) and depleting the amount in the trailing fractions (trough) (Fig.  4, A and B). Assuming that each CapG bound two PIP 2 molecules (see Table I), the K d obtained from five experiments performed with a range of CapG and/or PIP 2 concentration was 8.1 Ϯ 0.9 M (mean Ϯ S.E.). This is comparable with the spectroscopic titration result, affirming the validity of the two independent methods.
Gelsolin Binding to PIP 2 -Tryptophan titration could not be used to study gelsolin binding to PIP 2 because the full-length gelsolin signal (without phosphoinositide) fluctuated and did not reach a steady level even after 20 min. The reason for this instability was not investigated further. Gel filtration experiments showed that gelsolin binding to PIP 2 was enhanced by Ca 2ϩ (Fig. 5, A-F). At pH 7.5, the apparent K d values were 305.4 and 40.2 M with and without Ca 2ϩ (Table II). The latter value is similar to that of CapG, indicating that gelsolin and CapG have comparable PIP 2 binding affinity in the presence of Ca 2ϩ . However, in EGTA, gelsolin has a much higher K d than CapG, suggesting that Ca 2ϩ induces a larger change in binding affinity. This could be due to a disproportionate increase in k off relative to k on . 10 M Mg 2ϩ did not substitute for Ca 2ϩ (data not shown), consistent with previous results (46).
The effect of Ca 2ϩ was amplified when the pH was shifted from 7.5 to 7.0 (Fig. 5, compare A-C with D-F). The relations among K d , Ca 2ϩ , and pH are shown in Fig. 5G. In the absence of Ca 2ϩ , decreasing pH from 7.5 to 7.0 had minimal effect (K d of 300 and 350 M, respectively). This is not surprising, since PIP 2 protonation is not expected to change substantially within this narrow pH range (47) and a broader pH range does not affect binding of profilin to PIP 2 either (37). However, at pH 7.0, less Ca 2ϩ was required to increase binding. 0.2 M Ca 2ϩ decreased the K d by half at pH 7.0, while 4.5 M Ca 2ϩ was required to produce the same effect at pH 7.5. Both Ca 2ϩ concentrations are well within the range achieved following agonist stimulation, particularly at the cytoplasm immediately subjacent to the plasma membrane.
Ca 2ϩ and pH Regulation of Gelsolin Domains-To determine which part of gelsolin contributes to the Ca 2ϩ and/or pH dependence of PIP 2 binding, we examined the PIP 2 -binding characteristics of several gelsolin domains. Gelsolin contains six segmental repeats, S1-6 (32). The NH 2 -terminal half encompassing S1-3 binds actin independently of Ca 2ϩ (48) and has two known PIP 2 binding sites and potentially a third unmapped site (33,49,50). The COOH-terminal half (S4 -6), which requires Ca 2ϩ to bind actin (51), has not been examined previously for PIP 2 binding.
Unlike full-length gelsolin, the gelsolin NH 2 -terminal half behaved well during fluorescence titration (Fig. 6A). It bound PIP 2 with high affinity, and saturation was reached at a slightly lower PIP 2 concentration in EGTA than in Ca 2ϩ (the opposite of full-length gelsolin and CapG). The K d values for the experiment shown in Fig. 6A were 1.2 and 2.9 M, respectively. The stoichiometry of binding, derived from Fig. 6B, was 3.4. This value is twice that of CapG, confirming that gelsolin NH 2 -terminal half has more PIP 2 binding sites (33). Gel filtration studies confirmed that Ca 2ϩ increased the K d . The Hill coefficient of 1.1 Ϯ 0.03 (Fig. 6C, Table II) suggested that binding was noncooperative and that the sites bound PIP 2independently. S1, which has one PIP 2 site, bound 1.6 mol of  (Table II).
The gelsolin COOH-terminal half bound PIP 2 with much lower affinity (approximately 7-fold higher K d by fluorescence measurements) than the NH 2 -terminal half (Table II). It is therefore probably not involved in PIP 2 binding per se. As with the NH 2 -terminal half, binding to the COOH-terminal half was reduced in Ca 2ϩ (Fig. 7C). This is in sharp contrast to the large Ca 2ϩ -enhancement of PIP 2 binding to full-length gelsolin. The  opposite effects of Ca 2ϩ on full-length and half-length gelsolins therefore cannot simply be due to nonspecific lipid aggregation. The pronounced enhancement of PIP 2 binding to full-length gelsolin most likely reflects a Ca 2ϩ -dependent exposure of the NH 2 -terminal half PIP 2 binding sites through a change in the COOH-terminal half. This conclusion is based on the observation that neither the NH 2 -nor COOH-terminal halves are activated by Ca 2ϩ to bind PIP 2 , and only the COOH-terminal half is known to undergo Ca 2ϩ -induced conformational change (51). Gelsolin NH 2 -terminal half binding to PIP 2 was enhanced by lowering pH. The K d dropped from 8.2 to 3.4 M between pH 7.5 and 7.0 in the presence of EGTA (Fig. 7A). In contrast, the gelsolin COOH-terminal half was not affected by pH (Fig. 7B). DISCUSSION Actin polymerization in response to agonist activation is frequently associated with a rise in cytosolic Ca 2ϩ , changes in PIP 2 content, and intracellular pH. There is also compelling evidence that gelsolin, which severs and caps actin filaments in response to changes in Ca 2ϩ and PIP 2 concentration and pH, is involved in actin remodeling (17,(52)(53)(54). In this paper, we show that gelsolin and CapG binding to PIP 2 is affected by physiologically relevant changes in Ca 2ϩ and pH. The effects are not due to alterations in PIP 2 structure per se but reflect changes in the proteins. This is the first report that PIP 2 binding to any protein is directly modulated by signals generated during agonist stimulation and has implications for divergent PIP 2 -dependent processes beyond a direct effect on the cytoskeleton.
The finding that gelsolin binding to PIP 2 is promoted by Ca 2ϩ is consistent with the current model for how gelsolin is activated by Ca 2ϩ to bind actin (48,51). Our deletion studies suggest that the extreme COOH terminus of gelsolin is critical to the inhibition of the NH 2 -terminal actin binding sites, because gelsolin lacking the COOH-terminal 23 residues no longer requires Ca 2ϩ to bind actin (56). We do not know at present whether actin binding and PIP 2 binding are regulated identically. This question can now be addressed, because the actin and PIP 2 -binding sites of gelsolin have been mapped (33, 50, 56 -58) and the crystal structures of gelsolin S1 complexed with actin (57) and full-length gelsolin in EGTA 2 have been solved recently.
Less is known about how pH affects gelsolin conformation. Selve and Wegner (59) first reported that pH 6 increases the rate of gelsolin binding to actin in the presence of Ca 2ϩ . Lamb et al. (26) subsequently showed that the Ca 2ϩ requirement for gelsolin severing is reduced at pH 6.5 and abolished at pH below 6.0. pH 5 induces gelsolin unfolding, as determined by dynamic light scattering (26). We find that a less extreme pH drop potentiates Ca 2ϩ activation of PIP 2 binding to full-length gelsolin. Acidic pH increases the NH 2 -terminal half binding to PIP 2 even without Ca 2ϩ but has no effect on COOH-terminal half binding. Therefore, mild acidification probably promotes PIP 2 binding by directly altering the NH 2 -terminal PIP 2 binding sites.
The significance of an increase in PIP 2 affinity described here depends on the PIP 2 concentration in the plasma membrane. This is difficult to estimate precisely because PIP 2 may be partitioned and sequestered. One estimate, based on PIP 2 accounting for 1% of plasma membrane lipid, suggests that the PIP 2 concentration in the plasma membrane of a spherical cell with a radius of 10 m is 10 M (44). In platelets, the PIP 2 concentration is estimated to be about 300 M when averaged over the entire cell volume (internal and plasma membranes) (60), and PIP 2 concentration decreases by 30% following stimulation (16). Cytosolic [Ca 2ϩ ] rises during agonist stimulation, and the 4 -8-fold increase in CapG and gelsolin binding affinity described here is sufficiently large to promote their increased association with the plasma membrane despite a modest decrease in membrane PIP 2 . The magnitude of the increase depends on the PIP 2 concentration before and after stimulation. Immunogold labeling studies show that 4 and 6.5% of gelsolin is associated with the plasma membrane in resting and activated platelets, respectively (42). This represents a 63% increase in membrane association after stimulation. Our finding that Ca 2ϩ increases PIP 2 binding affinity can explain how PIP 2 uncaps gelsolin and CapG even as the plasma membrane PIP 2 content decreases following agonist stimulation.
Since only a handful of the currently identified PIP 2 -binding proteins are Ca 2ϩ -and pH-sensitive, our finding is consistent with a selective regulation of the gelsolin family. Nevertheless, increased gelsolin and CapG binding will impact multiple PIP 2dependent processes by altering PIP 2 availability to other binding proteins, especially when PIP 2 concentration is decreased during agonist stimulation. Some actin-binding proteins are inhibited by PIP 2 (profilin, cofilin, capping protein), while others are activated (␣-actinin and vinculin). Gelsolin and CapG can therefore exert positive as well as negative effects indirectly by controlling PIP 2 . We postulate that as the cytosolic [Ca 2ϩ ] rises during stimulation, gelsolin severs filaments and PIP 2 dissociates it from the filament end. Increased gelsolin binding to PIP 2 displaces capping protein and profilin, neither of which are Ca 2ϩ -sensitive, from the plasma membrane. Profilin catalyzes polymerization (16), and the reaction is terminated by capping protein-mediated filament capping (55). Multiple rounds of severing, uncapping, and facilitated actin addition at the barbed ends fuel explosive amplification of filament growth observed during lamellipodial extension and membrane ruffling.
Our findings also have implications beyond a direct effect on the cytoskeleton. Many important signaling proteins are regulated by PIP 2 as well. It is significant that several pleckstrin homology proteins (reviewed in Ref. 2) bind PIP 2 with similar affinity as the gelsolin family. For example, the K d values of ␤-adrenergic receptor kinase type 1, pleckstrin, dynamin, and phospholipase C ␦ are 50, 50, 4, and 1 M, respectively. Therefore, gelsolin and CapG can potentially compete with them for PIP 2 , particularly when the [Ca 2ϩ ] rises and PIP 2 level drops during agonist stimulation. This possibility is supported by in vitro and in vivo experiments. In vitro, gelsolin stimulates and inhibits inositol-specific phospholipase C isozymes in a biphasic manner (29). 3 Gelsolin stimulates phosphoinositide 3-OHkinase (31), although we find that gelsolin and CapG also inhibit it. 4 Gelsolin activates phospholipase D (30) in a PIP 2dependent manner. Modest overexpression of CapG (28) or gelsolin has profound effects on phospholipase C␤ and phospholipase C␥ activated through two distinct receptor-mediated pathways. 3 In conclusion, these observations show that gelsolin and CapG binding to PIP 2 is selectively regulated by second messengers. This regulation provides an additional level of control above that of a bulk change in PIP 2 content. Differential modulation and cross-talk between the PIP 2 -binding proteins allow control to be exerted at multiple points in the signaling cascade.